Ethical statement and study design
This study was approved by the Animal Care and Use Committee (approval number: BA1507-180/047-01) and carried out in accordance with the Guide for the Care and Use of Laboratory Animals of the Seoul National University College of Medicine (Republic of Korea). Total six shoulders from three adolescent mongrel dogs (approximately one year old and weighed 20 kg) were examined. Dogs were obtained from Kukje Laboratory Animal Center (Republic of Korea).
Experimental procedure and experimental animals
1. Gait analysis
In this study, we measured articular cartilage deformation patterns by using kinematics-based 3 dimensional (3D) shoulder models. We decided to use a marker-based motion tracking system (eight infrared cameras with a sampling frequency of 120 frames/sec, Eagle Digital real-time system, MotionAnalysis Co., USA). The large tracking areas of motion tracking systems are suitable for normal walking activities in in vivo canine models.
We decided to implant custom markers into shoulder bones rigidly to minimize skin artifacts in the motion tracking data. A custom marker was designed with three conventional reflective balls (diameter of 8 mm) (Fig 1a). Two custom markers were installed in the humerus and scapula using orthopedic stainless surgical threaded pins. Scapular pins were inserted in two locations along the scapular spine and humeral pins were inserted into the distal and proximal ends of the humerus along the lateral side of the diaphysis (Fig 1b). After four to five days of recovery following the marker insertion surgeries, we recorded the movements of the shoulder bones by tracking the ball markers as the dogs walked freely within our motion laboratory. Only one shoulder was tested in a single motion analysis examination to minimize the loosening of surgical pins. We found that the pins in shoulder bones were easily loosen when the tested dog laid the pin-inserted shoulder on the floor. Thus, the contralateral shoulder remained intact to allow the tested dog to use the intact shoulder in resting positions. After two weeks of recovery time, gait analysis of the contralateral shoulder was followed thorough the identical marker installation and motion tracking procedures.
2. 3D shoulder models
The 3D shoulder models were created from computed tomography (CT) (matrix: 512 × 512, field of view (FOV): 400 mm, slice thickness: 2.2 mm, slice spacing: 1 mm, Brilliance CT 64-channel, Philips, Netherlands) and magnetic resonance (MR) (T2 weighted fast field echo (FFE) sequence, repetition time (TR): 575 ms, echo time (TE): 11.51 ms, flip angle: 20°, matrix: 512 × 512, FOV: 130 mm, slice thickness: 2 mm, slice spacing: 2.2 mm, Achieva 3.0T TX, Philips, Netherlands) images of the shoulder bones of tested dogs. We tried to maintain the location and orientation of shoulder joints in each imaging step to minimize potential artifacts in MR images which might be generated from the magic angle effect . The shoulder bones and custom markers were automatically reconstructed from the CT images using the OSIRIX (Pixmeo, Switzerland) and GeoMagic (Research Triangle Park, NC, USA) software, while the articular cartilage layers were manually segmented from the MR images using a custom MATLAB code. The 3D bone and cartilage models were then combined by aligning the subchondral bone profiles of each model (best-fit alignment in GeoMagic software). The final shoulder models included scapular and humeral bones, articular cartilage, and custom markers (Fig 2a).
3. Supraspinatus tendon resection model
After completing articular cartilage contact strain and T2* relaxation time maps for intact shoulders, we completely resected the supraspinatus tendon in one of the shoulders while the contralateral shoulder remained intact. The resection area was covered with penrose drain tubes to prevent a recovery of the resected muscle. The dogs were allowed to perform normal activities in our animal research laboratory for three months following the resection surgeries (cage type: SUS304 stainless steel frame cage with fiber reinforced plastic (FRP) bottom plate (3.4 m × 1.3 m × 2.4 m), one dog per each cage with automatic watering system, temperature: 20 ± 2 °C, humidity: 50 ± 10%, light cycle (12 hours): 7 am to 7 pm). Measurements of the articular cartilage contact strain and T2* relaxation time patterns were then repeated for both shoulders in the supraspinatus-resected dogs. All surgical processes were conducted under general anesthesia with an intramuscular injection of 1) satropine sulfate (0.1 ml/kg of body weight, DAI HAN PHARM. Co., Ltd, Seoul, Korea) and 2) the mixture (0.2 ml/kg of body weight) of xylazine (Rompun, Bayer Korea, Seoul, Korea) and tiletamine-zolazepam (Zoletil 50, Virbac, Carros, France).
1. Gait pattern and articular cartilage strain measurement
The completed 3D shoulder models were aligned with the marker positions from the motion tracking data to determine the locations of shoulder bones in each gait frame (Fig 2a). The relative positions between the centers of the humeral head and glenoid cavity were calculated with six degrees of freedom (DOFs) (three translational and three rotational motions in a Cartesian coordinate system) (Fig 2b). An example of the rotational motion between the Y axis of the humeral head and glenoid cavity is shown in Fig 2b. We removed any abnormal gait cycles that were outside of the mean ± one standard deviation (red shaded area in Fig 2b) from all six DOF components. The remaining gait cycles were then averaged to generate a representative gait pattern for each dog (solid red line in Fig 2b).
Articular cartilage thickness was determined by calculating the perpendicular distance from the subchondral bone interface to the articular cartilage surface for both the scapular and humeral cartilage. Articular cartilage contact strain was defined as the ratio between the undeformed articular cartilage thickness and the thickness of the overlapping areas between the scapular and humeral articular cartilage in each gait fame (Fig 3a) [33, 34]. The articular cartilage contact areas were then defined as the cartilage areas in which the overlapping articular cartilage thickness was greater than 0.25 mm (in-plane resolution of the MR images). We created a 3D articular cartilage contact strain map for each gait frame from an entire representative gait cycle and combined all of these strain maps to generate a cumulative contact strain distribution. In this cumulative strain map, a cumulative contact area was the combination of individual contact areas from each gait frame.
2. Articular cartilage T2 star (T2*) relaxation time measurement
A qMRI scan of each tested dog was performed on the same day as the morphological CT and MR imaging. A T2 weighted multi-echo fast-field sequence (TR: 700 ms, TE: 3.83/9.37/14.91/20.46/26.01 ms, flip Angle: 25°, matrix: 768 × 768, FOV: 150 mm, slice thickness: 2 mm, slice spacing: 2.2 mm) was used in a 3T clinical MRI scanner (Achieva 3.0T TX, Philips, Netherlands). The T2* relaxation time of each pixel in the MR images was calculated using mono-exponential least squares fitting . T2* relaxation time values over 50 ms were removed to minimize the partial effects of synovial fluid around articular cartilage boundaries because T2 relaxation time of synovial fluid is known to be approximately 100 ms , and the T2* relaxation time is known to be around 50% of the T2 relaxation time values . We divided each articular cartilage surface into multiple 20° regions from the posterior to the anterior ends (Fig 3b) because our contact strain maps indicated anterior-to-posterior directional variations after the resection surgeries, but the changes in the medial-to-lateral direction were minimal.
3. Accuracy and reproducibility of contact strain measurements
Intra- and inter-observer reliability of manually segmented articular cartilage models was tested. Segmentation of articular cartilage surface and subchondral bone interface in a shoulder joint was done by three independent researchers in three difference days. Multiple 3 dimensional articular cartilage models were generated from various segmentation results for a shoulder joint. Articular cartilage thickness values at 70 random locations were compared in different cartilage models. Intra-class correlation coefficients (ICC) with 95% confident interval (CI) and average root-mean-square (RMS) differences in the thickness measurements among different segmentation results were calculated. We also calculated variations in average T2* relaxation time when articular cartilage thickness was increased or decreased within the average RMS difference in the thickness measurements.
Because articular cartilage contact strains were directly determined based on the location of the scapular and humeral bones, we decided to measure the accuracy and reproducibility of our motion-tracking-based 3D shoulder models using a custom plastic phantom model. The phantom consisted of two rectangular blocks with a plastic ball attached at one end of each block. The phantom blocks were positioned in three different configurations (relative angles of 0°, 30°, 60°). Two custom markers were installed in each phantom block to emulate the in vivo experiments (Fig 4). We moved the phantom model in the motion laboratory in random directions for 10 min while the motion tracking system continuously recorded the positions of the markers. We created a 3D model of the phantom blocks and measured 1) the center distance between the two phantom balls and 2) the angle between the two phantom blocks for 2000 randomly selected frames. The bias (average differences in the motion-tracking-based measurements from the physical measurements of the phantom model) and precision (variations in the motion-tracking-based measurements) of the center distance and angle measurements were then calculated to find systemic errors in motion tracking system. Finally, corresponding bias and precision of articular cartilage contact strain and contact area ratio (ratio between contact area and total cartilage surface area) in in vivo shoulder models were estimated by changing the position of scapular and humeral bones within the range of bias and precision values measured in the phantom experiments at each configuration (0°, 30°, 60° angles between scapula and humerus).
Intra- and inter-observer ICC with 95% confidence interval (CI) of segmentations were analyzed by using SPSS software (SPSS statistics 25, SPSS Inc., IL, USA). Pearson correlation coefficients between articular cartilage contact strain and T2* relaxation time were calculated by using MATLAB software (R2016a, MathWorks, MA, USA). A p-value less than 0.05 was considered to be a statistically significant difference.